Ectopic Pregnancy and Prior Induced Abortion
ANN ASCHENGRAU LEVIN, MS, STEPHEN C. SCHOENBAUM, MD, MPH, PHILLIP G. STUBBLEFIELD,
MD, SUSAN ZIMICKI, MS, RICHARD R. MONSON, MD, SCD, AND KENNETH J. RYAN, MD
Abstract: We compared the prior pregnancy his- control confounding factors reduced the relative risks
tories of 85 multigravid women with an ectopic preg- to 1.3 (95 per cent confidence interval, 0.6-2.7) and 2.6
nancy and 498 multigravid delivery comparison sub- (95 per cent confidence interval, 0.9-7.4), respective-
jects. We found a relationship between the number of ly. Theanalysis suggests that induced abortion may be
prior induced abortions and the risk of ectopic preg- one of several risk factors for ectopic pregnancy,
nancy: the crude relative risk of ectopic pregnancy particularly for women who have had abortions plus
was 1.6 for women with one prior induced abortion pelvic inflammatory disease or multiple abortions.
and 4.0 for women with two or more prior induced (Am J Public Health 1982; 72:253-256.)
abortions; however, use of multivariate techniques to
Introduction all cases using a standard questionnaire. Seven (4.0 per cent)
refused to participate and four (2.3 per cent) were discharged
An ectopic pregnancy can threaten a woman's life and before they could be reached for interview.
future fertility. Three previous investigations of the relation- We also selected approximately five comparison sub-
ship between induced abortion and subsequent ectopic preg- jects for each case. The comparison subjects were chosen
nancy have yielded conflicting results: a study from Greece randomly from the group of obstetrical patients who deliv-
found that women with positive abortion histories were ten ered on the day of the case's surgery. In all, we attempted to
times more likely to have an ectopic pregnancy than women interview 866 delivery patients. Of these, 28 (3.2 per cent)
without this history,' while Yugoslavian2 and Japanese3 refused to participate and 17 (2.0 per cent) were discharged
studies reported no excess risk for women with prior in- before they could be interviewed.
duced abortions. To investigate further this relationship The interview obtained information about past pregnan-
among women in the United States, we performed a case- cy outcomes, age, education, race, religion, payment meth-
control study comparing the obstetrical histories of women od, smoking habits, prior pelvic infections, pelvic surgery,
having an ectopic pregnancy with women having a term and contraceptive use. If the patient revealed a prior induced
delivery. abortion, details of the procedure including any complica-
tions were elicited. After the interview the subject's current
medical record and, whenever possible, the records of any
Methods admitted induced abortion were reviewed.
For the analysis, we included 171 cases who had patho-
Between July 1976 and May 1978, we identified all cases logical documentation of an ectopic pregnancy and five
at Boston Hospital for Women, Lying-In Division with a cases who met clinical criteria consistent with a tubal
final post-operative diagnosis of ectopic pregnancy (176 abortion. The latter had a positive test for the beta subunit of
women). Ninety-eight per cent of these pregnancies were human chorionic gonadotropin plus evidence of blood and/or
located in the fallopian tubes; two were ovarian, one was chorionic villi in the tube at surgery.
abdominal, and one was cervical. We attempted to interview First, we eliminated uninterviewed subjects (11 cases
and 45 comparison subjects) because we had no information
Address reprint requests to Kenneth J. Ryan, MD, Kate Macy regarding their obstetrical and medical history. Next we
Ladd Professor of Obstetrics and Gynecology, and Chairman of the eliminated 19 comparison subjects whose delivery occurred
Department of Ob-Gyn, Brigham and Women's Hospital, Harvard prior to 37 weeks' gestation so that all comparison subjects
Medical School, 75 Francis Street, Boston, MA 02115. Ms. Levin is had term deliveries. We also excluded primigravidas (61
Epidemiologist, Massachusetts General Hospital; Dr. Schoenbaum
is Assistant Professor of Medicine at the Brigham and Women's cases and 299 comparison subjects), subjects who became
Hospital, Harvard Medical School; Dr. Stubblefield is Assistant pregnant with an intrauterine device in place (31 cases and
Professor, Obstetrics and Gynecology, Harvard Medical School; five comparison subjects), subjects who had a prior steriliza-
Ms. Zimicki is assistant Scientist, International Center for Diarrheal tion (six cases and one comparison subject), and the three
Disease Research, Bangladesh; and Dr. Monson is Professor of
Epidemiology, Harvard School of Public Health. This paper, sub- repeat ectopic pregnancies of subjects who had already been
mitted to the Journal May 4, 1981, was revised and accepted for enrolled as a case and who became pregnant again during the
publication August 27, 1981. study period. Because some subjects fell into more than one
AJPH March 1982, Vol. 72, No. 3 253
LEVIN, ET AL.
TABLE 1-Percentage of Subjects According to Prior Repro- record included a history of one or more induced abortions
ductive Outcomes* omitted during the interview. In the analysis, women were
considered to have had an induced abortion if it was men-
Ectopic Comparison Odds of tioned either at the interview or in the medical record.
Cases Subjects Ectopic An examination of the obstetrical histories (Table 1)
Prior Outcome (N = 85) (N = 498) Pregnancy indicated that greater percentages of cases than comparison
Ectopic Pregnancy 14.1 0.8 20.32
subjects reported prior histories of ectopic pregnancy, in-
Induced Abortion 34.1 20.5 2.01 duced abortion, and spontaneous abortion. Fewer cases
Spontaneous Abortion 34.1 27.3 1.4 reported one or more prior livebirths.
Live Birth 67.1 79.1 0.51 The relative risk of ectopic pregnancy was 20.3 for
women with a prior ectopic pregnancy and 2.0 for women
1) p < 0.01 (Chi-square test) with one or more prior induced abortions. A history of prior
2) p < 0.001 (Chi-square test)
*The outcomes are not mutually exclusive; i.e., a subject may have had spontaneous abortion had little association with ectopic
more than one prior outcome. pregnancy (relative odds = 1.4); and a history of one or more
prior livebirths was associated with a reduced risk of subse-
quent ectopic pregnancy (relative odds = 0.5).
exclusion category, this left 85 cases and 498 comparison We reexamined these data omitting women with a prior
subjects for the final analysis. The relative risk, as approxi- ectopic pregnancy. Among the remaining subjects, the rela-
mated by the relative odds (odds ratio), was used to quanti- tive risk of an ectopic pregnancy following one or more
tate the association between the exposure (induced abortion) induced abortions was 2. 1, virtually the same as in the entire
and the outcome (ectopic pregnancy). series. In addition, for each of the four women who had had
The effect of confounding was assessed in two ways: we both a prior ectopic pregnancy and induced abortion, the
stratified the data by values of each potential confounder, induced abortion preceded the ectopic pregnancy.
calculated the Mantel-Haenszel summary relative risk,4 and There appeared to be a direct relationship between the
compared this risk to the crude relative risk; we used a number of prior induced abortions and the relative risk of
multiple logistic regression model5 including 15 variables to etopic pregnancy (Table 2). In comparison to women with no
assess the simultaneous effect of potential confounders. prior induced abortions, the crude relative risk of ectopic
Dichotomous variables included in the multivariate analysis pregnancy was 1.6 for women with a history of one induced
were patient's race (Black, other), religion (Catholic, other), abortion and 4.0 for women with a history of two or more
payment method (welfare and self-paying, other), parity induced abortions.
(zero, one or more), prior spontaneous loss, prior ectopic We next took into account the possible combined effects
pregnancy, prior use of oral contraceptives or intrauterine of the difference in pregnancy duration between the cases
devices, prior pelvic surgery, dilatation and curettage (unre- and comparison subjects and the selection of comparison
lated to an induced abortion), history of pelvic inflammatory subjects on the basis of a similar date of hospitalization to
disease, gonorrhea, and smoking. The regression analysis the cases. In other words, the cases were "at risk" for
also included two dichotomous variables for prior induced having an induced abortion for several months longer than
abortion, one for one induced abortion and another for two the comparison subjects. However, when we analyzed a
or more induced abortions. We simultaneously entered all subset of the data matching the subjects on the date of their
variables into the model and calculated the standardized last menstrual period, the increased risk for cases remained
relative risk for each variable while controlling for the other unchanged.
variables. The distribution of induced abortion methods as report-
ed by the women was similar among the cases and compari-
son subjects (Table 3). Almost twice as many of the prior
Results induced abortions among the cases were performed in a
foreign country or as an illegal procedure in the United
Twenty-nine cases (34 per cent) and 102 comparison States. When women with "foreign" or illegal abortions
subjects (20 per cent) reported one or more induced abor- were excluded from the analysis, the relative risks of ectopic
tions at interview or had a history of induced abortion in pregnancy given a history of one and two or more prior
their medical records. For 76 per cent of these cases and 56 induced abortions were 1.3 and 3.3 respectively.
per cent of these comparison subjects, there was complete
agreement between the interview and medical record. For 17 TABLE 2-Percentage of Subjects According to Number of
per cent of the cases and 26 per cent of the comparsion Prior Induced Abortions
subjects, the medical records either contained no informa- Prior Ectopic Comparison Crude
tion regarding prior pregnancy outcome or did not specify Induced Cases Subjects Relative
whether a prior abortion was induced or spontaneous. For Abortion(s) (N = 85) (N = 498) Odds
another 7 per cent of the cases and 10 per cent of the
0 65.9 79.5 1.0
comparison subjects, the medical record omitted induced 1 23.5 17.3 1.6
abortions revealed at interview. Finally, for 19 per cent of 2+ 10.6 3.2 4.0
the comparison subjects, but for no cases, the medical
254 AJPH March 1982, Vol. 72, No. 3
ECTOPIC PREGNANCY AND PRIOR ABORTION
TABLE 3-Percentage of Subjects with a Prior Induced Abor- apy following pregnancy, a history of infertility and pelvic
tion According to Characteristics of (Last) Induced surgery. Pathological evidence of tubal infection (e.g.,
Abortion chronic salpingitis, follicular salpingitis or chronic follicular
Ectopic Comparison salpingitis) was found in 63.5 per cent of the cases. The
Cases Subjects crude relative risks of ectopic pregnancy for women with
Characteristics (N = 29) (N = 83) prior pelvic surgery, gonorrhea, and prior pelvic infection
unrelated to induced abortion were 2.3, 5.7, and 11.1 respec-
Abortion Method tively (Table 4).
Vacuum Aspiration 69.0 65.1
Dilatation and Curettage 20.7 28.9 We examined the extent to which these and other
Other (Intrauterine Injection, variables confounded the relationship between prior induced
Hysterotomy) 10.3 6.0 abortion and subsequent ectopic pregnancy. Standardization
Illegal or "Foreign" Procedure* 31.0 15.7 using the Mantel-Haenszel technique revealed that, individ-
ually, most variables including race, payment method, prior
Curettage 10.3 4.9 pill and IUD use, infertility, prior spontaneous loss, dilata-
Bleeding 1+ weeks 51.7 41.9 tion and curettage, history of pelvic infection (either omitting
Pain 1+ weeks 10.3 4.0 or including post-abortal infections), and gonorrhea ex-
Fever 17.2 6.2 plained little of the association between the exposure and
Infection 17.2 3.6 outcome. Only when the data were standardized for parity or
*Induced abortion was performed in a foreign country or as an illegal for smoking was the crude relative risk of ectopic pregnancy
procedure in the United States. following one or two or more prior induced abortions
reduced by at least 15 per cent.
Using the multiple logistic regression model including
The ectopic cases reported post-abortal complications the 15 variables from Table 5, the adjusted relative risk of
more frequently than did the comparison subjects. These ectopic pregnancy was 1.3 for women with one prior induced
complications included an additional post-abortal curettage, abortion (95 per cent confidence interval 0.6-2.7) and 2.6 for
extended bleeding and pain, fever and infection. Post-abortal women with two or more prior induced abortions (95 per
infections were reported four to five times more frequently cent confidence interval 0.9-7.4). The most important risk
by the cases. Moreover, 60 per cent (3/5) of the cases vs 33 factors for ectopic pregnancy were prior ectopic pregnancy
per cent (1/3) of the comparison subjects experienced post- (RR = 7.7), prior pelvic infection (RR = 7.5), prior pelvic
abortal infections after an illegal induced abortion. Howev- surgery (RR = 2.6), and payment method (RR = 3.0).
er, since only a small number of women (five cases and three Excluding post-abortal infections, the crude relative
comparison subjects) actually had post-abortal infections, risks of ectopic pregnancy for women having one or more
their exclusion from the analysis did not alter the results. prior induced abortions and for women having a prior history
A variety of gynecological problems, more common of pelvic infection were 1.5 and 9.1, respectively. There
among the cases, included: prior pelvic and gonorrheal appeared to be a synergistic relationship between these two
infection (excluding post-abortal infections), antibiotic ther- exposures and the occurrence of ectopic pregnancy: Eleven
TABLE 5-Standardized* Relative Risks of Ectopic Pregnancy
TABLE 4-Percentage of Subjects According to Selected Gyne- and 95 Per Cent Confidence Intervals According to
cologic Characteristics Selected Characteristics
Comparison 95 Per Cent
Ectopic Cases Subjects Relative Confidence
Characteristic (N = 85) (N = 498) Characteristic Risk Interval
History of History of one induced abortion 1.3 0.6- 2.7
Pelvic infection* 29.4 3.6 History of two or more induced abortions 2.6 0.9- 7.4
Gonorrhea 16.5 3.2 History of ectopic pregnancy 7.7 1.9-31.5
Antibiotic therapy History of pelvic infection 7.5 3.5-16.0
following a pregnancy 37.6 20.1 Payment method 3.0 1.5- 6.0
Medical assistance to History of pelvic surgery 2.6 1.4- 4.6
become pregnant 22.4 13.9 History of gonorrhea 2.5 0.9- 7.1
Prior pelvic surgery 52.9 28.9 Prior dilatation and curettage 1.6 0.8- 3.3
Prior endometriosis 3.5 3.4 Smoking 1.5 0.8- 2.8
Congenital abnormality Prior IUD use 1.4 0.7- 2.5
of tubes or uterus 0.0 1.2 Race 1.4 0.6- 3.0
Clomiphene induced Prior spontaneous loss 0.9 0.4- 1.9
pregnancy (current) 2.4 1.2 Religion 0.8 0.4- 1.5
>12 mos. to get Prior pill use 0.6 0.3- 1.2
pregnant (current) 38.7 7.8 Parity 0.5 0.2- 1.0
*Excluding post-abortal infections *Standardized using the multiple logistic regression model.
AJPH March 1982, Vol. 72, No. 3 255
LEVIN, ET AL.
of the women having an ectopic pregnancy vs only four of educational level, age, and gravidity; however, they did not
the delivery patients had a prior history of both exposures, assess the data for confounding by prior gynecological
yielding a crude relative risk of 25.0, which greatly exceeds infection or surgery. Induced abortion is illegal in Greece,
the sum of the risks for either factor alone (10.6). The and infection appears to be more common after illegal
increased risk remained after controlling for prior IUD use abortion. The effects of post abortal infection are too impor-
and age. tant to be ignored.
Finally, data from a Japanese study3 indicate that one or
more prior induced abortions were associated with a crude
Discussion relative risk of 2.4 for subsequent ectopic pregnancy. Ad-c
justing for hospital and year of ectopic pregnancy, the
There are many risk factors for ectopic pregnancy and relative risk fell to 1.3. These data are similar to our results
several are indicated in the results of this study. When we for persons who had one prior induced abortion.
used mnultivariate techniques to control the effects of these There are several possible explanations for the fact that
factors simultaneously, there was no detectable increase in the existing studies are not in complete agreement; the most
the risk of ectopic pregnancy for women who had had one likely are the differences in study design and analytic meth-
prior induced abortion. The risk for women who had had two ods just described. Differences in the populations, such as
or more prior abortions fell from an initial estimate of 4.0 to prior exposure to pelvic infections, or differences in abortion
2.6, and was no longer statistically significant. While we procedures, such as variations in the experience of the
cannot eliminate ch-ance as an explanation of our findings, a surgeons or in the type of instruments used, could account
possible association of multiple prior induced abortions with for some of the variation in results and would be extremely
subsequent ectopic pregnancy persists. difficult factors to measure. The present study clearly indi-
The Ljubljana Abortion Study found that neither one cates that there are several risk factors for ectopic pregnancy
nor two or more induced abortions altered the crude relative which are interrelated and it suggests that under certain
risk of ectopic pregnancy.2 Stratifying by age, the authors circumstances prior induced abortion(s) may increase the
compared the obstetrical histories of incident cases of ectop- risk of subsequent ectopic pregnancy. The existence of a
ic pregnancy and a control group designed to represent a large number of risk factors for ectopic pregnancy means
population of intrauterine pregnancies. Deiivery patients and that future studies directed at elucidating the precise role of
women seeking induced abortions were pooled into the induced abortion will have to be extremely large in order to
control group in a ratio reflecting the author's obstetrical control confounding.
practice. It is hard to justify the inclusion of controls who are
seeking an induced abortion, since their frequency of prior
abortion is comparatively high, or primigravidas, whose REFERENCES
1. Panayotou PP, Kaskarelis DB, Miettinen OS, Trichopoulos DB,
frequency of prior induced abortion is, by definition, zero. Kalandidi AK: Induced abortion and ectopic pregnancy. Am J
When we re-analyzed their data and compared multigravid Obstet Gynecol 1972; 114:507-510.
cases and multigravid delivery controls, the crude relative 2. Hren M, Tomazevic T, Seigel D: Ectopic pregnancy. In: Andol-
risks of ectopic pregnancy for Yugoslavian women with one sek L (ed): The Ljubjana Abortion Study, 1971-1973. Bethesda:
and two or more prior induced abortions were 1.7 (p < 0.05) National Institutes of Health, 1974; 34-38.
3. Sawazaki C, Tanaka S: The relationship between artificial abor-
and 2.4 (p < 0.025), respectively. These crude relative risks tion and extrauterine pregnancy. In: Harmful Effects of Induced
are similar in trend to our crude risk estimates. Furthermore, Abortion. Tokyo: Family Planning Federation of Japan, 1966;
while the authors found that a history of pelvic inflammatory 49-63.
disease, sterility, and gynecological treatment also increased 4. Mantel N, Haenszel W: Statistical aspects of the analysis of data
from retrospective studies of disease. J Natl Cancer Inst 1959;
the risk of ectopic pregnancy, they did not perform a 22:719-748.
stratified or multivariate analysis to assess confounding and 5. Prentice RL: Use of the logistic model in retrospective studies.
interaction by these variables. Biometrics 1976; 32:599-606.
A study from Greece, found a tenfold increase in the
risk of ectopic pregnancy among women wtih prior induced ACKNOWLEDGMENTS
abortions and the increased risk was present for both single This study was supported by a grant from the National Founda-
and multiple aborters. The authors matched for hospital, tion-March of Dimes.
256 AJPH March 1982, Vol. 72, No. 3